Ultrasonic Deposition of High-Selectivity Nanoporous Carbon Membranes

See allHide authors and affiliations

Science  17 Sep 1999:
Vol. 285, Issue 5435, pp. 1902-1905
DOI: 10.1126/science.285.5435.1902


Ultrasonic deposition creates a thin film of polymer on a tubular, macroporous, stainless steel support. Using polyfurfuryl alcohol as the nanoporous carbon precursor and a pyrolysis temperature of 723 kelvin, a membrane was prepared with the following permeances, measured in moles per square meter per Pascal per second: nitrogen, 1.8 × 10−12; oxygen, 5.6 × 10−11; helium, 3.3 × 10−10; and hydrogen, 6.1 × 10−10. The ideal separation factors as compared to that for nitrogen are 30:1, 178:1, and 331:1 for oxygen, helium, and hydrogen, respectively.

Molecular separations are energy-intensive processes that can be used in a wide spectrum of areas, ranging from emerging biotechnologies to more classical fuels and chemicals production. The global move toward greener production is necessarily coupled with the need for more energy-efficient, and hence novel, separations. Molecular separations based on differences in molecular size and shape using ceramic membranes (1,2) at ambient temperature are particularly intriguing, because these materials can conduct such separations with less energy. This approach is particularly important for the removal of water from biotechnological broths in which the hydraulic load is high, but it is just as much an issue for gas separations such as the separation of nitrogen (N2) from air, which is done by energy-intensive cryogenic distillation. Ceramic, as opposed to polymeric, membranes can operate under harsh conditions and at elevated temperatures (3), such as occur in reactive separations. Zeolites (4), sol-gels (5), and nanoporous carbons (NPCs) (6) are each candidate materials for the preparation of novel size-selective ceramic membranes.

Continuous nanoporous carbon membranes can be prepared as thin films supported on porous stainless steel by means of ultrasonic deposition (UD) of polyfurfuryl alcohol (PFA). These supported nanoporous carbon membranes (SNPCMs) show unprecedented size and shape selectivities, even among very small molecules with similar dimensions. Using the rate of N2 permeation for comparison, hydrogen (H2), helium (He), and oxygen (O2) are transported through an SNPCM 330, 178, and 30 times faster. If the molecules were transported on the basis of Knudsen diffusion rather than on the basis of size selectivity, these transport ratios (H2/N2, He/N2, and O2/N2) would be 3.73, 2.65, and 0.94, respectively. The ultrasonic deposition technique makes it possible to prepare membranes as films that are mechanically robust and selective but remain below a critical thickness for crack formation.

Membrane synthesis and fabrication have proven to be critical problems with these materials, because continuous films are difficult to achieve. Nanoporous carbon molecular sieves are disordered solids with densities between 70 and 80% of that of crystalline graphite. Nanopore dimensions range from 0.3 to 0.7 nm, with a maximum between 0.35 and 0.55 nm (7–10). These structural features, which are reminiscent of fullerenelike fragments, make the NPC molecular sieves ideally suited for the production of N2 from air by kinetic separation (8,9). Membranes consisting of NPC on support media offer a new and potentially more efficacious means to O2/N2and other small-molecule separations (6).

Two types of NPC membranes have been prepared in the past: unsupported (planar monoliths and hollow fibers) and supported (asymmetric membranes) (11). The unsupported NPC materials provided high selectivities but suffered from extreme fragility. Recent work avoided this drawback by forming NPC layers on support media such as porous graphite and sintered stainless steel (6). These materials were more robust and displayed good selectivity but reproducibility was problematic. Polymer precursor deposition and pyrolysis are two of the key steps that must be precisely controlled to overcome this problem.

The membrane supports that we used were sintered stainless steel tubes (SS304, Mott Metallurgical) with an outside diameter of 6.35 mm, a wall thickness of 1.48 mm, and a nominal pore size of 0.2 μm, and were coated with PFA resin with an ultrasonic nozzle (12). The low-momentum deposition provided by an ultrasonic nozzle offers more precise control of the coating process as compared with conventional high-pressure gas spraying, including a 102 to 103 lower spray velocity that minimizes penetration into the support. The droplets are narrowly distributed in size from 10 to 100 μm, depending on nozzle operating frequency. The precursor delivery rate can be accurately controlled with a syringe pump. The coated tubes were then pyrolyzed (13), and upon completion of the heat treatment the furnace was shut off and the tubes were allowed to cool.

We prepared four membranes (SNPCM I through IV) whose total masses and NPC layer thicknesses are presented in Table 1. Typically, the first carbon layer ranged in mass from 8 to 11 mg, corresponding to a carbon yield of 20 to 30%, based on the wet coat mass. In successive steps after each heat treatment, we added less NPC per coat: Second coat masses were 5 to 7 mg, and third coat masses were 0.1 to 0.4 mg. In the case of SNPCM-IV, however, three primary coatings with a total mass of 5 mg were put down, followed by three more coatings of 1.4, 1.0, and 1.1 mg.

Table 1

Permeances and ideal separation factors for SNPCMs. The pyrolysis temperature for each sample is given. Thicknesses are calculated on the basis of final masses. Samples I through III were coated three times, and sample IV was coated six times.

View this table:

Scanning electron microscope (SEM, Hitachi S-4000) images were taken of the exterior surface and of cross and axial sections of a sacrificial membrane. We prepared this membrane under the same conditions as those used for SNPCM-II. The micrograph (Fig. 1) reveals a film with a uniform radial thickness of 15 ± 2 μm. The axial profiles were similarly uniform, and the surface view showed a continuous defect-free coating.

Figure 1

An SEM radial image of the membrane surface.

In order to measure the gas permeance, the tubular membranes were inserted into a module that consisted of a cylindrical membrane holder with knife-edge flanges on either side sealed with copper gaskets. The flanged ends were welded to compression fittings. With external compression applied to the fittings, pressures up to 7000 kPa were maintained, with no measurable leakage at 295 K. Pressure-rise experiments were performed on all tubular membranes (14).

The N2 permeance of another sacrificial membrane prepared in the same way as SNPCM-III was measured after the addition of each NPC layer. The N2 permeance initially decreased as the NPC mass increased (Fig. 2A). Once a critical mass of NPC was reached, the last carbon layer added resulted in an increase in permeance and a concomitant loss in selectivity. An SEM analysis of this membrane film revealed that catastrophic crack formation had occurred (Fig. 2B). In this case, the total mass of NPC on the support was 14.8 mg before the cracking process and 15.5 mg after. Given that the density of the NPC is 1.6 g/cm3(9), these masses correspond to calculated thicknesses of 18.2 and 19.1 μm, respectively. Asymmetric films of this kind are known to have a critical thickness below which the film has mechanical properties that are more plasticlike than those of the bulk material (15). Based on these data, we can estimate that the onset of bulk-like NPC properties and catastrophic cracking occurs in the vicinity of a critical film thickness of 20 ± 3 μm. Thus, the deposition process must be carefully controlled in order to produce membranes below this thickness.

Figure 2

(A) N2 permeance as a function of carbon mass. (B) An SEM image of a cracked surface of a film 19.1 μm in thickness.

We sought to prepare the membranes (SNPCM-I through IV) so that the final masses of NPC produced films that were below the critical thickness. SNPCM-III consisted of three layers of NPC with corresponding masses of 11.0, 5.9, and 0.4 mg and calculated thicknesses of 13.5, 7.3, and 0.5 μm. The change in shell side pressure as a function of time for each gas shows that the slopes of the pressure-rise curves decrease with increasing molecular size (Fig. 3). There is a measurable difference between the slopes of these rates even for N2 and O2, which differ by only 0.02 nm in kinetic diameter (16). The small-molecule selectivity can be explained by examining the pore size distribution for a similar SNPCM prepared at 873 K with a mean pore size centered around an average of 0.50 nm. (Fig. 4). From the ideal gas law, the time rate of change in shell side pressure,P ss, is (17)Embedded Image(1)Integrating from the initial shell side pressure P ss0 at timet 0 = 0 to the final shell side pressureP ss at t f =t results inEmbedded Image(2)A plot of the left side of Eq. 2 versus time gives a straight line with a slope ofEmbedded Imagethe gas permeance. Permeances and ideal separation factors (the ratio of permeance with respect to N2 permeance) for SNPCM-III (Table 1) are 30 to 90 times greater than the calculated separation factors based on Knudsen diffusivities (16).

Figure 3

Pressure rise as a function of time for sample SNPCM-III.

Figure 4

Pore size distribution for a SNPCM.

It was assumed in the integration of Eq. 1 that π0was independent of pressure. To test this assumption, the permeances were measured as a function of pressure from 300 to 7000 kPa and were found to be pressure-independent (Fig. 5). This result shows that there was no Poiseville flow through the membrane.

Figure 5

Permeance as a function of pressure for sample SNPCM-III.

Finally, using UD to control thickness, the effect of membrane synthesis temperature during pyrolysis was tested by also preparing materials at 423, 573, and 873 K (Table 1). The membrane prepared at 423 K was impermeable, whereas that prepared at 873 K displayed permeabilities that were over an order of magnitude greater than those of the others. Its selectivities were less than those of SNPCM-III but were still higher than the Knudsen values. These selectivities indicate that the membrane was not cracked but that the higher pyrolysis temperature brought on lower selectivity. When pyrolized at 573 K, the resulting membrane, SNPCM-II, showed good selectivities but not as high as those of the membrane prepared at 723 K. That thermal effects should play such a role in determining membrane properties, and hence selectivities, is expected on the basis of previous work with bulk NPC materials (9). O2/N2 selectivity versus O2permeability for SNPCM-II, -III, -IV, and two additional membranes prepared at 723 K are plotted versus the best known reported values (18), which lie on or below the dividing line (Fig. 6).

Figure 6

O2/N2 selectivity versus O2 permeability (π′).

In addition to transient experiments, a steady-state experiment was run using SNPCM-III to separate air. High-pressure air from a gas cylinder was fed at a constant pressure of 3550 kPa to the core side of the membrane. The shell side operated at atmospheric pressure, and a gas chromatograph was used to analyze the permeate, which contained 44% oxygen, a 100% enrichment.


View Abstract

Navigate This Article